Complete Guide to Thinking Machines Inkling

-->

Complete Guide to Thinking Machines Inkling - Analytics Vidhya

India's Most Futuristic AI Conference Is Back – Bigger, Sharper, Bolder

d

:

h

:

m

:

s

Career

GenAI

Prompt Engg

ChatGPT

LLM

Langchain

RAG

AI Agents

Machine Learning

Deep Learning

GenAI Tools

LLMOps

Python

NLP

SQL

AIML Projects

Reading list

How to Become a Data Analyst in 2025: A Complete RoadMap

A Comprehensive Learning Path to Tableau in 2025

A Comprehensive NLP Learning Path 2025

Learning Path to Become a Data Scientist in 2025

Step-by-Step Roadmap to Become a Data Engineer in 2025

A Comprehensive MLOps Learning Path: 2025 Edition

Roadmap to Become an AI Engineer in 2025

A Comprehensive Learning Path to Master Computer Vision in 2025

Best Roadmap to Learn Generative AI in 2025

GenAI Roadmap for Enterprises

Large Language Models Demystified: A Beginner’s Roadmap

Learning Path to Become a Prompt Engineering Specialist

Complete Guide to Thinking Machines Inkling

Harsh Mishra Last Updated : 20 Jul, 2026

8 min read

Thinking Machines Lab has unveiled Inkling, its first general-purpose open-weights foundation model. It is a multimodal MoE model with 975B parameters, 41B active parameters, and a 1M-token context window. Rather than chasing benchmark supremacy, Inkling is designed as a customizable foundation for multimodal reasoning, agentic AI, coding, tool use, audio and vision tasks, and domain-specific fine-tuning. In this article, we explore Inkling’s architecture, training, capabilities, benchmarks, pricing, deployment options, fine-tuning workflow, and enterprise readiness.

Table of contents

What is Thinking Machines Inkling?

Inkling Architecture

How Inkling Processes Text, Images, and Audio

Training and Post-Training

Self-Fine-Tuning

Inkling Benchmark Results

How to Access Inkling?

Option 1: Tinker Playground

Option 2: Hugging Face

Option 3: Hosted Inference Providers

Option 4: Self-Hosted Inference

Let’s Try Thinking Machine’s Inkling

Task 1: Test Inkling in the Playground

Task 2: Prompt for Tool Use

Inkling-Small

Conclusion

Frequently Asked Questions

What is Thinking Machines Inkling?

Inkling is a general-purpose, multimodal, open-weights AI model from Thinking Machines Lab. It processes text, images, and audio, and generates text outputs. Released on July 15, 2026, Inkling is available under the Apache 2.0 license, allowing commercial use, modification, and redistribution. Rather than offering a fixed API-only model, Thinking Machines designed Inkling to be customized and fine-tuned for specific domains, workflows, and enterprise needs.

Inkling at a Glance

Property Inkling Specification

Developer Thinking Machines Lab

Model type Decoder-only multimodal Transformer

Architecture Sparse Mixture-of-Experts

Total parameters 975 billion

Active parameters 41 billion

Transformer layers 66

Routed experts 256

Shared experts 2

Experts selected per token 6 routed experts plus 2 shared experts

Maximum model context Up to 1 million tokens

Tinker context options 64K and 256K

Input modalities Text, image, audio

Output modality Text

Training tokens 45 trillion

Weight formats BF16 and NVFP4

License Apache 2.0

Fine-tuning platform Tinker

Local inference support Transformers, SGLang, vLLM, TokenSpeed, llama.cpp and related tools

Hosted providers Together AI, Fireworks, Modal, Databricks and Baseten

Although Inkling has 975B parameters, it activates only about 41B per token. This reduces compute costs compared to a dense model of the same size, though all expert weights still need to be stored across the serving infrastructure.

Inkling Architecture

Inkling is a 66-layer, decoder-only, multimodal Mixture-of-Experts (MoE) Transformer. Its high-level architecture is shown below:

Sparse Mixture-of-Experts Backbone

Each MoE layer contains 256 routed experts and 2 shared experts. For every token, the router activates 6 routed experts, while the 2 shared experts remain active throughout. Routing is based on sigmoid scores with auxiliary-loss-free load balancing. The selected experts’ outputs are then normalized and combined. While inspired by DeepSeek-V3’s MoE design, Inkling extends it with its own multimodal processing, hybrid attention, positional encoding, convolutional layers, and post-training optimizations.

Hybrid Local and Global Attention

Inkling uses five sliding-window attention layers followed by one global attention layer. Local layers reduce computation by focusing on nearby tokens, while every sixth layer enables information exchange across the wider context. The model also uses eight key-value heads.

Relative Positional Embeddings

Instead of RoPE, Inkling uses learned relative positional representations. These representations model the distance between query and key tokens, helping the model handle relationships across long sequences. Thinking Machines reports that this approach showed better long-context extrapolation in its experiments.

Short Convolutions

Inkling applies short convolutions:

After key and value projections

Before attention and MLP outputs rejoin the residual stream

These operations help capture short-range patterns, while attention handles broader relationships.

Multi-Token Prediction

Inkling includes multi-token prediction layers that draft several future tokens at once.

During speculative decoding, the main model verifies these drafts in parallel. Accepted tokens are generated faster, while incorrect drafts are regenerated, improving inference speed without changing the final output distribution.

How Inkling Processes Text, Images, and Audio

Inkling converts all supported modalities into representations that enter the same decoder.

Modality How Inkling Processes It

Text Text is tokenized and processed autoregressively, like other decoder-only language models.

Images Images are split into 40×40 patches and encoded using a four-layer hierarchical MLP. Recommended image size: 40-4096 pixels.

Audio Audio is converted into dMel (discretized mel-spectrogram) embeddings. For best results, use 16 kHz WAV audio under 20 minutes. Tinker also supports WAV, MP3, and FLAC.

Video Video was included during pretraining, but out-of-the-box video capabilities were not formally evaluated. Fine-tuning is recommended for production use.

Training and Post-Training

Inkling was pretrained on 45 trillion tokens spanning text, images, audio, and video. Its training data combined public, third-party, synthetic, and augmented sources, with extensive cleaning, deduplication, quality filtering, and safety processing.

Hybrid optimization

Thinking Machines used:

Muon for large matrix weights

Adam for other parameters

Learning-rate-dependent weight-decay scheduling

The company reports that coupling weight decay to the square of the learning rate helped keep weight magnitudes stable across different training horizons.

Supervised fine-tuning

The initial post-training bootstrap used synthetic supervised data generated by several open-weights models, including Kimi K2.5. The company describes this supervised stage as a small portion of total post-training compute.

Reinforcement learning

Most post-training compute was reportedly spent on large-scale reinforcement learning across:

Mathematics

Reasoning

Agentic coding

Tool use

Vision

Audio

Conversation

Instruction following

Calibration

Safety

Thinking Machines scaled asynchronous reinforcement learning beyond 30 million rollouts. Its held-out aggregate reasoning reward increased from 0.264 after SFT initialization to 0.356 for the released checkpoint.

Aggregate reasoning reward increased log-linearly across more than 30 million reinforcement-learning rollouts. | Source: Thinking Machine

Training hardware

Inkling was trained on NVIDIA GB300 NVL72 systems. This was Thinking Machines Lab’s first major foundation-model training effort.

Self-Fine-Tuning

A notable new capability demonstrated at launch is Inkling’s ability to help fine-tune itself through coding assistants and the Tinker platform.

In the demonstration, Inkling was asked to become a lipogram assistant that avoids using the letter “e.” Using a coding-agent workflow, it:

Defined the target behaviour

Generated synthetic training examples

Wrote evaluation logic

Prepared and launched a Tinker fine-tuning run

Evaluated the updated checkpoint

Loaded the fine-tuned model back into the agent environment

This does not mean the model changes its own weights autonomously during a normal conversation. Instead, Inkling can use coding tools to design, execute, and evaluate its own customization workflow. This makes model fine-tuning more accessible and reduces the amount of manual engineering required.

Inkling Benchmark Results

Inkling is built as a balanced general-purpose model, delivering strong performance across reasoning, coding, agentic workflows, instruction following, vision, audio, factuality, and safety instead of optimizing for a single benchmark. Its standout capabilities include mathematical reasoning, visual reasoning, audio understanding, instruction following, and agentic coding, although it does not consistently outperform the best open or closed models in every category.

Overall Capability Profile

Source: Thinking Machine

Official benchmarks were evaluated with a reasoning effort of 0.99, temperature 1.0, and a maximum coding trajectory of 256K tokens. Since benchmark outcomes depend on evaluation settings, tools, and execution environments, they should be viewed as comparative indicators rather than absolute rankings. The radar chart below highlights Inkling’s strongest results on AIME 2026, GPQA Diamond, IFBench, VoiceBench, and visual reasoning benchmarks, while also revealing weaker performance on factuality and long-horizon agentic coding tasks.

Reasoning, Coding, and Agentic Performance

Inkling performs well on mathematical and scientific reasoning, with strong results on AIME 2026 and GPQA Diamond. It also benefits significantly from tool use, retrieval, and external computation for complex reasoning tasks.

For coding, Inkling handles repository-level tasks, tool calling, terminal interactions, and multi-step workflows, making it one of the stronger open-weights models. While leading closed models still outperform it on benchmarks like SWE-bench Pro and Terminal Bench, Inkling stands out by combining agentic coding, multimodal inputs, a 1M-token context window, open weights, and fine-tuning flexibility.

Impact of Reasoning Effort

Source: Thinking Machine

Inkling allows developers to adjust reasoning effort based on task complexity. Higher reasoning effort generally improves performance, but the gains diminish at higher token budgets. Different benchmarks benefit to varying degrees, so the optimal setting is the lowest effort that consistently meets your quality target.

Factuality and Instruction Following

Inkling performs well on instruction-following tasks, making it suitable for structured workflows involving formatting rules, tool use, and multi-step instructions. However, factuality remains a weakness, especially for tasks requiring precise knowledge recall.

For high-stakes applications such as healthcare, legal, finance, and enterprise search, Inkling should be paired with retrieval, source citations, verification tools, and human review.

Vision and Audio Performance

Native support for images and audio is one of Inkling’s biggest strengths. It can understand charts, diagrams, documents, mathematical visuals, spoken instructions, and long audio recordings without requiring separate models.

Source: Thinking Machine

Inkling delivers strong vision and audio performance, with competitive results on AudioMC, MMAU, and VoiceBench. Its visual reasoning further improves when combined with Python tools, highlighting the benefits of integrating perception with external computation.

Safety Performance

Thinking Machines evaluat

[truncated for AI cost control]